Modeller

Build powerful, predictive models....

Modeller’s flexible design allows its users to develop powerful, predictive models for use in all stages of the credit cycle including PD, LGD and EAD models through a clear, easy-to-navigate interface. The software is the product of Paragon’s 25 years of experience in the credit industry – experience that gives our customers absolute confidence in our tools and is why Modeller is used across the lending industry.

Modeller2.png

What Modeller has to offer...

 

Flexible

Modeller’s flexible scorecards accommodate themselves to the needs of your business rather than forcing your business to conform to the needs of the scorecards.

Interactive

The Modeller development interface allows “hands on” experience for Credit Managers to analyse and evaluate all facets of credit risk.  Modeller requires little training to take full advantage of its many capabilities.

Adaptable

Modeller is adaptable to the requirements of many different users. The standard scorecard building interface provides familiarity and repeatability while the advanced process flow interface allows advanced users to add many additional functions

Cost-Effective

Modeller quickly pays back good returns on your investment, enhances the efficiency of your risk management operations and adds repeatability and audit to your scorecard building process

No Black Boxes

There are no hidden processes in Modeller. Paragon will be happy to explain the functionality in any of their model building and scoring techniques

Fully Supported

If you have any questions/queries about Modeller whilst you are developing your scorecard our dedicated support team are here to help. Onsite training is also provided 

Features

"A fully integrated, predictive model development system encompassing the entire Modelling life-cycle"

 

User-friendly Interface


Modeller provides a user-friendly point-and-click interface for all modelling tools. This enables experienced users to take advantage of the faster throughput while new team members can quickly engage with the model building process through the use of the simplified ‘Overlay’ view.

 

Variable Classification

Modeller’s “Autogroup” function automates the classification process giving the user a statistically sound starting point for manual group manipulation. Modeller’s dynamic point-and-click classification interface enables the user to quickly and easily adjust group boundaries whilst the group information (e.g. WoE) and the univariate information (e.g. Gini) is updated on-screen in real-time

Advanced Modelling Techniques


Modeller caters for a wide-range of regression techniques including:

  • Logistic Regression

  • Elastic Net Models

  • Survival Analysis (Cox PH)

  • Decision Trees

  • Continuous outcome prediction for ancillary functions (e.g. LGD)

Elastic Net Small.jpg

Variable Selection and Dimensionality Reduction

Field selection is one of the most time consuming processes in scorecard building. Modeller’s “Field Reducer” tool uses an automated process to perform this task, removing the ‘junk’ based on user-defined parameters giving the user more time to concentrate on other stages of the development process

 

Contact Us...

If you have any queries please  just get in touch...

Thanks! Message sent.

Features

"A fully integrated, predictive model development system encompassing the entire Modelling life-cycle"

User-friendly Interface


Modeller provides a user-friendly point-and-click interface for all modelling tools. This enables experienced users to take advantage of the faster throughput while new team members can quickly engage with the model building process through the use of the simplified ‘Overlay’ view.

 

Variable Classification

Modeller’s “Autogroup” function automates the classification process giving the user a statistically sound starting point for manual group manipulation. Modeller’s dynamic point-and-click classification interface enables the user to quickly and easily adjust group boundaries whilst the group information (e.g. WoE) and the univariate information (e.g. Gini) is updated on-screen in real-time

Advanced Modelling Techniques


Modeller caters for a wide-range of regression techniques including:

  • Logistic Regression

  • Elastic Net Models

  • Survival Analysis (Cox PH)

  • Decision Trees

  • Continuous outcome prediction for ancillary functions (e.g. LGD)

Variable Selection and Dimensionality Reduction

Field selection is one of the most time consuming processes in scorecard building. Modeller’s “Field Reducer” tool uses an automated process to perform this task, removing the ‘junk’ based on user-defined parameters giving the user more time to concentrate on other stages of the development process

Elastic Net.jpg

Expert Modelling Techniques

When there are insufficient defaulting accounts to use statistical methods and, in particular, when there are no defaults, we employ expert modelling techniques linked to customer data. This lets us take account of the correlations between the predictor characteristics and to eases the path to a full statistical model by applying Bayesian update techniques.

 

Low Volume (WoE Approximation) Techniques

Providing a link between expert models and full statistical approaches, Paragon Low Volume Models enable the estimation of default rates within univariate characteristics while retaining the characteristic relationships required to restrain model overfitting.

Model Validation


Modeller has an advanced reporting tool which enables the user to produce parametrised or ad-hoc reports. These can be used to justify the validity of scorecards and their potential benefits. A number of report templates are available or you can design custom reports from scratch using our flexible report designer tool

Documentation

Modeller comes with an Office add-in that is compatible with Word and Excel. This add-in enables the documentation of the entire scorecard development directly into a Word document or Excel spreadsheet

Audit

 

There are no black-box techniques in Paragon Modeller, all models are fully audited and our procedures are open to scrutiny for testing and reproducibility.

 

Modeller’s audit trail is a complete record of the model building process on a practically button press by button press basis​

Deployment

Modeller automatically includes full coding instructions for each model to ease the process of IT deployment. Models built in Modeller can be exported in multiple formats for implementation across various different platforms (SAS, Java, PMML). Customers with advanced decision engines such as Paragon’s Paris system can deploy models directly without IT involvement.

Sharing Resources

Modeller promotes collaboration between team members by allowing sharing of data definitions, groupings and report templates.

Paragon_TRANSPARENT.png